Model-based point cloud alignment with principle component analysis for robot welding

In this paper, we propose a fast pose estimate method to estimate the relative pose between a measured point cloud and a reference point cloud, in order to increase the speed and accuracy of the point cloud registration procedure. The method extracts feature points from the point cloud and find the principal components of the features points' coordinates. The relative pose of the measured point cloud is found by minimizing the principal components of the measured point cloud and that of the reference point cloud. After aligning the measured point cloud with the reference point, we apply the Iterative Closed Point registration procedure to find the final match. The method can find the rotation and transformation relationship of two similar point clouds in arbitrary positions in a short time. It can be applied to determine the positions of another similar target in the case of a given object pose.